JP2009218743A - Ip protocol processor and its processing method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To accelerate the reassembling processing of fragmented IP packets and to reduce a memory cost required for parallel processing. <P>SOLUTION: The IP protocol processor reassembles the IP packets on the basis of a bit table which manages the reception conditions of the fragmented IP packets. When reassembling them, the bit table to be the object of the reassembly is transferred from a first storage part storing a plurality of bit tables to a second storage part storing a part of the plurality of bit tables. On the basis of the bit table transferred to the second storage part, the fragmented IP packets are reassembled. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to protocol processing in an application device that uses IP (Internet protocol) communication and an IP communication control device, and more particularly, to a technique for receiving an IP packet that is transmitted by being divided into a plurality of fragment packets. Further, it is suitable for an embedded device that realizes speeding up of IP protocol processing by a TOE (TCP / IP Offload Engine) technology.

  IP reassembly is one of the processes with a large amount of processor processing among protocol processes. In IP communication, there is a mechanism in which one IP packet is divided and transferred to a plurality of IP packets on the transmission source or transmission path, and one packet of the division source is restored and received on the receiving side. The process of dividing one IP packet into a plurality of IP packets and transferring it is called an IP fragment (IP Fragmentation). On the other hand, the process of receiving a plurality of divided IP packets and restoring them to the original IP packet is called IP reassembly. The specification of IP reassembly is disclosed in RFC791.

  For IP reassembly, two algorithms, a processing method described in RFC 791 and a processing method described in RFC 815, are known. Since the present invention relates to the former method, the former method will be introduced. Here, the former IP reassembly processing method is referred to as a bitmap table method.

  In IP reassembly using the bitmap table method, the receiving side prepares a bit table in which 8 octets of payload data of a source IP packet are assigned to 1 bit. Since the IP datagram has a maximum length of 65535 octets, the prepared bit table needs to be about 8 kilobits long. Then, while storing the payload data of each arriving fragment packet, by setting the bit table bits corresponding to the offset position and length of the fragment data, the reception status of the fragment data is managed. After that, when all the bits on the bit table corresponding to the payload length of the source IP packet are set, it means that the data necessary for reassembly of the source IP packet has been prepared. Is completed.

  In recent years, network communication has rapidly increased in speed, and Ethernet (registered trademark) products compatible with gigabit networks have already become widespread for consumer use. Furthermore, a standard capable of realizing a transmission rate of 40 megabits / second or 100 megabits / second is now being standardized in the IEEE 802.3 working group. As network communication speeds up, the amount of data transmitted and received by applications using IP communication has also increased, and the load on communication protocol processing has increased for communication terminal devices.

  Conventionally, protocol processing for IP communication is often realized by software called TCP / IP (Transmission Control Protocol / Internet Protocol) protocol stack. Even in an embedded device having an IP communication function, IP protocol processing is often implemented by software. However, since a small-scale embedded device is often equipped with a CPU having a low operating frequency, which is advantageous in terms of manufacturing cost and power consumption, a processor load for protocol processing by necessary software increases.

  For this reason, it has become difficult to achieve high communication performance with conventional processing. Another problem is that CPU resources cannot be sufficiently allocated to application processing due to processing load on communication processing.

  As a countermeasure against such a problem, there is a technology collectively called TOE (TCP / IP Offload Engine) that reduces the communication processing load of the processor and realizes high-speed IP communication. This TOE is a technique for speeding up TCP / IP protocol processing by processing means different from a processor that executes an application.

In the TOE, the whole or part of the protocol processing is executed by hardware processing (integrated circuit) to increase the speed. In addition, in embedded devices, mounting is performed in which a TOE is converted into an LSI and incorporated in a chip.
JP 2007-274056 A

  In the above-described bitmap table type IP reassembly, the table size required to correspond to the maximum packet size of the IP packet before division is 1 Kbyte (8192 bytes). In addition, the same number of bitmap tables as the number of IP parallel reassembly processes is required. When the number of received IP packets per unit time increases due to higher communication speed, the IP reassembly that is processed in parallel also increases, and the memory size required for the bitmap table increases.

  On the other hand, it is desirable to form a bit table on an on-chip memory with a small access delay in order to realize high speed IP reassembly by hardware processing by the above-described TOE technology. However, a large memory capacity corresponding to the number that can be processed in parallel is required, and an on-chip memory having a high memory cost increases the memory cost. On the other hand, there is a problem that a limited number of parallel processing is possible in a small-capacity memory.

  As a prior art for managing the arrival status of fragment data using this bit data, for example, Patent Document 1 is known. In Patent Literature 1, by limiting the fragment size of an IP fragment to a narrow range, an implicit fragment number is assigned to the fragment packet, and the management of the reception status of the fragment packet is simplified and speeded up.

  In this method, since the size for dividing the IP packet is limited in advance, the bit width for managing the reception status of the fragment packet can be made smaller than 8K bits, so the memory size required for IP reassembly Can be reduced. However, since the division size is limited to a narrow range, it is not suitable for performing IP reassembly of fragment packets of any size.

  An object of the present invention is to speed up the reassembly process of fragmented IP packets and reduce the memory cost for parallel processing.

  The present invention is an IP protocol processing apparatus for reassembling the IP packet based on a bit table for managing the reception status of fragmented IP packets, and a first storage means for storing a plurality of bit tables; A second storage unit storing a part of the plurality of bit tables; and the reassembling from the first storage unit to the second storage unit when performing the reassembly based on the bit table. Transfer means for transferring the bit table to be subject to the transfer, and processing means for reassembling the fragmented IP packet based on the bit table transferred to the second storage means by the transfer means. It is characterized by.

  The present invention further includes a first storage unit that stores a plurality of bit tables and a second storage unit that stores a part of the plurality of bit tables. A processing method of an IP protocol processing device for reassembling the IP packet based on a bit table to be managed, wherein when performing the reassembling based on the bit table, a second storage unit A transfer step of transferring the bit table to be reassembled to the storage means, and a process of reassembling the fragmented IP packet based on the bit table transferred to the second storage means by the transfer means And a process.

  According to the present invention, the reassembling process of fragmented IP packets can be accelerated, and the memory cost for parallel processing can be reduced.

  The best mode for carrying out the invention will be described below in detail with reference to the drawings.

[First Embodiment]
FIG. 1 is a block diagram illustrating an example of a configuration of a network communication apparatus according to the first embodiment. As shown in FIG. 1, a CPU 102, a ROM 103, and a RAM 104 are connected to the system bus 101. Here, the CPU 102 reads the system program stored in the ROM 103 into the RAM 104 and executes it. The RAM 104 is a main storage device used when executing the system program.

  Reference numeral 105 denotes an IP protocol processing apparatus according to the present invention, which is connected to the system bus 101. The IP protocol processing device 105 executes TCP / IP protocol processing necessary for application communication executed by the CPU 102 and data transmission / reception to / from the network 117.

  In the IP protocol processing device 105, the bus bridge circuit 107 connects the internal local bus 106 to the system bus 101 and can transfer data between the local bus 106 and the system bus 101. The IP protocol processing device 105 and an external system composed of the CPU 102, the ROM 104, and the RAM 104 are connected to each other so that communication data and control data can be input and output by bus transfer. ing.

  Further, two local processors 108 and 109 for executing TCP / IP protocol processing, a local RAM 110, and a DMA controller (DMAC) 111 for transferring data are connected to the local bus 106. Further, a reassembly bitmap controller (RBMC) 112 for executing IP reassembly bit table processing and a communication timer 113 used for protocol processing timer processing are connected to the local bus 106. Further, a CAM (Content Addressable Memory) 114 which is an associative memory for storing and retrieving various management information for protocol processing is connected to the local bus 106. Further, an encryption processing unit 115 that executes encryption / decryption calculation processing necessary for encrypted communication processing, and a communication control unit 116 that transmits and receives data to and from the network 117 are connected to the local bus 106.

  The IP protocol processing device 105 executes IP protocol processing by parallel processing using a plurality of processor processes. Furthermore, using each hardware processing unit 111 to 115, TCP / IP protocol packet transmission / reception, transmission flow control, congestion control, communication error control, and encryption communication protocol processing such as IPsec and SSL / TLS are executed. .

  The programs executed by the local processors 108 and 109 are stored in the ROM 103 and are copied onto the local RAM 109 when the IP protocol processing apparatus 105 is activated. The local processors 108 and 109 read the program from the local RAM 109 and execute it.

  In FIG. 1, two local processors 108 and 109 are shown, but the number of processors that execute software processing is not limited to two.

  The DMAC 111 performs transfer processing of transmission / reception data and control data between memory modules and hardware modules inside or outside the IP protocol processing apparatus 105. The RBMC 112 performs bit manipulation processing of a bit table used in a bitmap (RFC 791) IP reassembly algorithm, and transfers bit table data to and from the RAM 104.

  The communication timer 113 is used for various timeout processing in IP protocol processing, such as measurement of timeout time in IP reassembly, time measurement required for various timers such as retransmission processing in TCP protocol communication, and transmission of acknowledgment response. . The encryption processing unit 115 executes encryption / decryption calculation processing with a heavy processing load in encryption communication protocol processing such as IPSec and SSL (Secure Socket Layer) / TLS (Transport Layer Security).

  The communication control unit 116 is equipped with functions for performing MAC processing (transmission media control processing) of the network 117 and transmission / reception of frame data. The network 117 is a wired network such as Ethernet (registered trademark), but may be a wireless network, an optical fiber network, or the like.

  Here, an example of an implementation configuration of the RBMC 112 will be described with reference to FIG. FIG. 2 is a block diagram illustrating an example of the configuration of the RBMC in the first embodiment.

  The RBMC 112 includes a register 201, a resource management unit 202, an internal memory (SRAM) 203, a bit processing unit 204, and a DMA function unit 205. First, the register 201 is a control interface for setting and starting the RBMC 112, and is a plurality of register sets. The local processors 108 and 109 read from and write to the register 201 to set and control the RBMC 112.

  The resource management unit 202 performs data management of a plurality of bit tables. The management information data is stored in the local RAM 109 shown in FIG. The bit processing unit 204 performs bit operation processing on an arbitrary range of bit table data stored in the internal memory 203. In this bit manipulation processing, all bits are set to 1 for a bit string of arbitrary length, all bits are cleared to 0, the number of bits having a value of 1 is counted, and when all bits are set to 1, 0 to 1 Contains a count of the number of bits changed.

  In this example, the internal memory 203 is implemented as a memory having a very small access delay. The RBMC 112 temporarily stores bit table data on which the bit processing unit 204 performs bit operation processing on the internal memory 203. In the bit operation processing of the bit processing unit 204, data is read from and written to the internal memory 203.

  The DMA function unit 205 is controlled by the resource management unit 202 and executes bit table data transfer between the internal memory 203 and a memory outside the RBMC 112.

  In the above configuration, an IP reassembly bit table is secured in the RAM 104. When the RBMC 112 performs bit manipulation processing of the bit table, the target bit table data is temporarily stored in the internal memory 203, and the bit manipulation processing is executed on the temporarily stored bit table data.

  Here, the bit manipulation processing of the IP reassembly bit table will be described with reference to FIG. In the bit table used in IP reassembly, each bit is mapped every 8 octets from the top of the payload of the source IP packet in order from the top bit. That is, each bit is a flag for storing whether or not the corresponding 8 octets have been received by IP reassembly. For example, in the bit table 301 shown in FIG. 3, 8 octets corresponding to a bit having a value of 1 indicate that it has already been received.

  In IP reassembly, the fragment offset and fragment size in the payload of the IP packet before division can be obtained from the contents of the IP header of the received fragment packet for the fragment data carried by the fragment packet. Based on this information, each bit in the bit range corresponding to the received fragment data is written to 1 indicating reception completion. When the last fragment packet is received, the total length of the source IP packet can be obtained, and the necessary bit width is determined on the bit table. Therefore, when all the bit strings corresponding to the payload length of the segmentation source IP packet are filled with the value 1, all fragment data constituting the segmentation source IP packet are received.

  In the example shown in FIG. 3, reference numeral 301 denotes a bit table, where 303 represents an offset position in the fragmented IP packet payload of the received fragment data, and 304 represents a bit width corresponding to the fragment data. Here, the bit table is a table for managing the reception status of fragmented IP packets.

  When the bit processing unit 204 receives the fragment data, the bit processing unit 204 writes 1 to each bit by the width of 304 from the position of 303. As a result, the bit data 301 is updated as indicated by 302. A portion represented by 305 represents a bit newly changed from 0 to 1. The bit processing unit 204 executes such bit operation processing.

  The bit processing unit 204 counts the number of bits whose value has changed from 0 to 1, and sets the newly received fragment data size in the register 201 of the RBMC 112.

  As described above, the bit processing unit 204 performs bit manipulation processing on data on the internal memory 203 with a small access delay. As a result, it is possible to increase the speed of IP reassembly, and therefore it is conceivable that the internal memory 203 is mounted with an SRAM, for example.

  However, placing all the IP reassembly bit tables for parallel processing in the internal memory 203 with a small access delay requires a memory capacity corresponding to the number of IP reassembly that can be processed in parallel, and the memory cost is very high. turn into.

  Therefore, all the bit tables are secured in the RAM 104, and only the bit table data to be processed by IP reassembly is copied to the internal memory 203 and temporarily stored, so that the memory cost of the internal memory 203 can be suppressed.

  In the first embodiment, an identifier called an IP reassembly ID is assigned to each IP reassembly to be processed in parallel, and managed in association with a bit table. Of the bit tables on the RAM 104, a bit table having a high frequency of bit operation processing is temporarily stored in the internal memory 203.

  FIG. 4 is a diagram showing a state in which the IP reassembly bit table is managed by the IP reassembly identifier. In this example, N bit tables that are being processed are secured on the RAM 104 at a certain point in time, and four bit tables are temporarily stored in the internal memory 203.

  As shown in FIG. 4, in the RAM 104, N bit tables indicated by 405 to 411 are arranged at arbitrary memory addresses. Among these, bit tables 406, 408, 409, and 411 having IP reassembly IDs of 1, 3, M, and N−1 are temporarily stored in locations indicated by 401, 402, 403, and 404 in the internal memory 203, respectively. Yes.

  That is, the method for storing data in the internal memory 203 is a full associative method with the bit table size as a unit. In the usage example of the internal memory 203 shown in FIG. 4, a bit table having IP reassembly IDs of 1, 3, M, and N−1 is placed in the internal memory 203 at a certain time. Therefore, if the next fragment packet received from that point is one of these IP reassembles, there is no need to transfer the bit table from the RAM 104 to the internal memory 203.

  However, when an IP reassembly fragment packet that does not store a bit table in the internal memory 203 is received, it is necessary to copy the bit table to be processed from the RAM 104 to the internal memory 203. Further, for example, when there is no space for storing the bit table in the internal memory 203, it is necessary to write back one of the temporarily stored bit tables to the RAM 104 and temporarily store the bit table to be processed next in the vacant place. Occurs.

  For this reason, the resource management unit 202 in the RBMC 112 manages the bit table for each reassembly ID. The bit table management information 70 that the resource management unit 202 holds for each IP reassembly ID includes the following 701 to 710. Reference numeral 701 denotes a reassembly ID. Reference numeral 702 denotes a record of the processing time of the IP reassembly processed last in the IP reassembly using this bit table. Reference numeral 703 denotes a record of the time interval (arrival interval) of the IP reassembly processed last. Reference numeral 704 denotes a memory address of the bit table in the RAM 104. Reference numeral 705 denotes a flag indicating whether or not the bit table is temporarily stored in the internal memory 203. Parameters 706 to 710 are used when the bit table is temporarily stored in the internal memory 203.

  Reference numeral 706 denotes an address of the bit table data on the internal memory 203, and reference numeral 707 denotes the size of the bit table data on the internal memory 203. A flag 708 indicates whether the bit table data has been changed in the internal memory 203. Reference numeral 709 denotes an offset address indicating the start position of the changed bit range of the bit table, and reference numeral 710 denotes a data size indicating the changed range. The management information of each bit table is stored in the local RAM 110 and updated by the resource management unit 202 of the RBMC 112.

  When the bit management data is transferred from the RAM 104 and stored in the internal memory 203, the resource management unit 202 checks the bit table management information to determine whether or not there is a space for temporarily storing the bit table data. If there is no free space, the last processed IP reassembly time interval 703 in the bit table management information is compared, and the IP reassembly bit table having the longest arrival interval is selected. For the selected bit table, data is written back from the internal memory 203 to the RAM 104, and an empty area is created in the internal memory 203.

  In this way, even when the bit table is not temporarily stored in the internal memory 203 by IP reassembly, the bit table data is transferred from the RAM 104 to the internal memory 203 and temporarily stored.

  Note that data transfer between the internal memory 203 and the RAM 104 is executed by the DMA function unit 205 in the RBMC 112.

  Next, the overall processing flow of IP reassembly executed by the IP protocol processing device 105 will be described with reference to FIG. FIG. 5 is a diagram showing a temporal processing sequence of the DMAC 111, the RBMC 112, and the local processor 108 in IP reassembly. Note that the example shown in FIG. 5 is a sequence on the assumption that the received IP packet is a fragment packet.

  When the communication control unit 116 receives the IP packet, first, the DMAC 111 starts transferring the received IP packet in S501, and the transfer is performed from the communication control unit 116 to the RAM 104. In addition, data having a size corresponding to the portion from the top of the received IP packet to the IP header is transferred to the RAM 104 and simultaneously transferred to the local RAM 110. When the transfer of the IP packet is completed in S502 up to the data size corresponding to the head MAC header (link layer frame header) and the IP header portion, a reception notification is sent to the local processor 108 (S503). Further, after the notification, the DMAC 111 continues to transfer the payload portion in S504.

  On the other hand, the local processor 108 receives the notification from the DMAC 111 in step S505, and starts IP packet reception processing. That is, the local processor 108 starts the IP packet reception process without waiting for the entire received IP packet to be transferred to the RAM 104. Next, the local processor 108 analyzes the contents of the MAC header and the IP header transferred to the local RAM 110 in S506. Here, it is checked whether the destination MAC address and the destination IP address can be received. In step S506, it is determined whether the packet is a fragment packet based on the contents of the IP packet header.

  Next, when it is determined in S506 that the packet is a fragment packet, in S507, the IP reassembly ID reassembly ID search process for the received fragment packet is performed. Subsequently, preloading of the bit table corresponding to the reassembly ID searched in S507 is activated for the RBMC 112 in S508 (S509). This preload is a process of temporarily storing the bit table data to be processed in the internal memory 203.

  In response to the activation instruction, the RBMC 112 checks whether the bit table of the reassembly ID is stored in the internal memory 203 in S510, and if not stored, temporarily stores the bit table secured in the RAM 104. That is, the bit table data is transferred between the RAM 104 and the internal memory 203 in S510.

  On the other hand, after starting the preloading of the RBMC 112, the local processor 108 performs verification of the IP header checksum of the received IP packet (that is, the fragment packet) in S511 without waiting for the completion of the processing. If the checksum of the IP header is correct, it is determined that the IP packet is receivable because the content of the IP header is correct.

  Next, in S512, the bit manipulation to the bit table is started for the RBMC 112 (S513). Then, the local processor 108 waits for completion of the bit operation processing in S514.

  In response to the activation instruction, the RBMC 112 performs bit manipulation processing on the bit table data temporarily stored in the internal memory 203 in S514. In this process, the data size corresponding to the number of bits whose bit value is changed from 0 to 1, that is, the newly received fragment size is set in the register 201, and the process completion is notified (S515).

  Here, when the local processor 108 receives the notification, the newly received fragment data size can be obtained by the IP reassembly process.

  On the other hand, when the DMAC 111 finishes transferring the entire received IP packet to the RAM 104 in S516, it notifies the local processor 108 of the transfer completion (S517). As a result, when the local processor 108 receives the DMA transfer completion notification in S518, the local processor 108 ends the IP packet reception process.

  The above is the overall processing flow of IP reassembly executed by the IP protocol processing device 105.

  In the first embodiment, it is determined in step S511 whether the received IP packet can be received after the IP header checksum verification. However, the RBMC 112 does not store the bit table data in the internal memory 203 at the stage of the bit manipulation processing of the IP reassembly bit table in S514 after the determination. In step S508, the bit table preload is activated in the RBMC 112. That is, while the local processor 108 is verifying the IP header checksum in S511, the RBMC 112 executes preload processing in S510.

  With such a processing sequence, the IP packet reception processing of the local processor 108 and the preload processing of the RBMC 112 are executed in parallel. As a result, even when data transfer occurs between the internal memory 203 and the RAM 104 when the RBMC 112 temporarily stores the bit table data in the internal memory 203, the time overhead can be ignored.

  As a result of S511, if an IP packet can be received, IP reassembly is executed in S512. Here, since the bit table data is already stored in the internal memory 203, the bit manipulation processing in S514 is executed at high speed. Even if it is determined in S511 that the IP packet cannot be received, it is clear that the bit table state is not changed because only the bit table data is secured in the internal memory 203 in S510.

  Next, the processing of the local processor 108 executed in S506 to S808 in the above-described IP reassembly processing sequence will be described in detail with reference to FIGS.

  FIG. 6 is a diagram showing a format of an IP header of an IPv4 (IP version 4) packet. Whether or not the IP packet is a fragment packet is determined by the fragment continuation flag 606 and the fragment offset 609. If the fragment continuation flag 606 is 1, it indicates that this IP packet is a fragment packet, and that there is another IP packet carrying subsequent fragment data.

  If the fragment continuation flag 606 is 0 but the fragment offset 609 is not 0, this IP packet is a fragment packet carrying the end of fragment data. With such a determination method, it is determined whether the received IP packet is a fragment packet.

  In addition, fragment packets constituting the same fragment source packet must have the same four field values of the IP identifier 605, the protocol 611, the source IP address 613, and the destination IP address 614 in the IP header.

  In the first embodiment, search entry data is stored in the CAM 114 by associating these four parameters with a reassembly ID for identifying IP reassembly. The search entry data is saved when a new IP reassembly is started and a new reassembly ID is determined. Also, the IP reassembly search process when a fragment packet is received is to search the CAM 114 for a reassembly ID using four fields obtained from the IP header as search keys. If no reassembling ID is found as a result of this search, a new IP reassembly is started and a new IP reassembly is determined.

  FIG. 7 is a flowchart showing the processing of the local processor 108 (S505 to S508 in FIG. 5). First, in S701, the MAC header and various parameters of the IP header of the received IP packet are read and acquired. In step S702, it is determined whether the destination MAC address and the destination IP address can be received. As a result of the determination, if the address cannot be received, the received packet is discarded, and this process ends.

  If the destination MAC address and the destination IP address are both receivable addresses, the process proceeds to S703. In S703, it is checked whether or not the IP packet is a fragment packet based on the values of the fragment continuation flag 606 and the fragment offset 609 in the IP header acquired in S701. If the packet is not a fragment packet, normal IP packet reception processing is executed without executing IP reassembly.

  On the other hand, if it is determined in S703 that the packet is a fragment packet, the process proceeds to S704, and a search process is executed to obtain the reassembly ID of the fragment packet. Since this search process is as described above, its description is omitted.

  Subsequently, in S705, it is determined whether or not the reassembly ID can be searched. If the reassembly ID cannot be searched, the process proceeds to S706. In S706, since the fragment packet needs to start a new IP reassembly, a new reassembly ID is determined. Then, a search entry is registered in the CAM 114 by associating a new reassembly ID with the four values of the packet IP identifier, protocol, source IP address, and destination IP address. Furthermore, a bitmap table corresponding to the new reassembly ID is secured on the RAM 104. In the subsequent S707, the reassembly ID newly determined in S706 and the memory address of the bit table secured on the RAM 104 are set in the register 201 of the RBMC 112. At this time, the resource management unit 202 of the RBMC 112 creates new IP reassembly bit table management information in the local RAM 110.

  If the reassembly ID can be searched in S705 and if the process in S707 is completed, the process proceeds to S708. In S708, the preload processing is started by setting the reassembly ID, the fragment offset 609 acquired from the IP header, the fragment data size that is the length of the payload of the IP packet, and the register 201 of the RBMC 112.

  Next, the bit table preload processing of the RBMC 112 executed in S510 in the above-described IP reassembly processing sequence will be described with reference to FIG.

  FIG. 8 is a flowchart showing bit table preload processing. First, in step S <b> 801, it is checked whether a bit table corresponding to the reassembly ID designated in the register 201 is temporarily stored in the internal memory 203. Here, if it is stored in the internal memory 203, this processing is terminated.

  If the bit table of the specified reassembly ID is not stored in the internal memory 203, the process proceeds to S802, and it is checked whether or not there is a free space for storing the bit table in the internal memory 203. If there is a vacancy, the process proceeds to S807, and if there is no vacancy, the process proceeds to S803.

  In step S803, for the IP reassembly in which the bit table is stored in the internal memory 203, the time interval from the previous process when each was last processed is compared, and the IP reassembly with the longest time interval is selected. Here, referring to the bit table management information 70 held by the resource management unit 202 for each IP reassembly ID, the time of the IP reassembly last processed in the bit table management information for which the internal memory storage flag 705 is on. Comparison is made with the value of the interval 703. That is, the reassembly ID having the largest value 703 in the bit table management information is selected.

  In step S804, it is checked whether the bit table on the internal memory 203 of the selected reassembly ID has been changed. Here, it is checked whether or not the internal memory data change flag 708 is on in the bit table management information of the selected reassembly ID. If changed, the process proceeds to S805, and if not, the process proceeds to S807.

  In step S <b> 805, only the range in which the bit table on the internal memory 203 of the selected reassembly ID is changed is written back to the bit table on the RAM 104. The RBMC 112 updates the data change offset 709 and the data change size 710 of the bit table management information during the bit manipulation process. Therefore, the changed portion of the bit table that needs to be written back is in the range of 709 to 710. Here, only the changed part of the bit table on the internal memory 203 is written and reflected in the bit table on the RAM 104.

  Next, in S806, all fields 705 to 710 of the management information are cleared for the bit table written back to the RAM 104, and the process proceeds to S807.

  In S807, the bit table data of the designated reassembly ID is read into the internal memory 203. The read location on the internal memory 203 is the location found in S802 or the location of the reassembly ID bit table selected in S803.

  In step S808, the bit table data in the range of the fragment offset and the fragment data size set in the register 201 is preferentially transferred to the internal memory 203 first. Bit table data other than this part is transferred later. This is to enable the bit manipulation process to be performed immediately when the bit manipulation process is activated before the entire bit table is transferred to the internal memory 203. The bit table preload process is executed as described above.

  Next, the bit manipulation processing of the bit table of the RBMC 112 executed in S514 in the above-described IP reassembly processing sequence will be described with reference to FIG.

  FIG. 9 is a flowchart showing bit manipulation processing of the bit table. First, in S901, the current timer value is acquired from the communication timer 113 to obtain the current time. Also, a difference value from the processing time 702 of the IP reassembly processed last in the bit table management information of the specified reassembling ID is obtained, and the time interval with the IP fragment packet that has been processed one time before is acquired. . Then, the fields 702 and 703 of the management information are updated.

  In step S <b> 902, it is checked whether the bit table of the designated reassembly ID is temporarily stored in the internal memory 203. Here, the resource management unit 202 of the RBMC 112 checks whether the internal memory storage flag 705 of the bit table management information of the reassembly ID is on. If it is on, the process proceeds to S906, and if it is off, the process proceeds to S903. Proceed to the process.

  In S903, a processing mode is performed in which a bit operation is directly performed on the bit table in the RAM 104. In step S904, the fragment offset specified in the register 201 and the bit range corresponding to the fragment data size are set to 1 for the bit table on the RAM 104. In S905, the number of bits changed from 0 to 1 in the bit table on the RAM 104 is set in the register 201, and the process proceeds to S910.

  On the other hand, in S906, the bit table data on the internal memory 203 is set to a bit operation mode. Next, in S907, the fragment offset specified in the register 201 and the bit range corresponding to the fragment data size are set to 1 for the bit table on the internal memory 203. In step S <b> 908, the number of bits changed from 0 to 1 in the bit table on the RAM 104 is set in the register 201. In step S909, the data offset and size of the portion changed in the bit table on the internal memory 203 are recorded in the bit table management information 709 and 710, respectively. Then, the process proceeds to S910.

  In S910, the end of the bit manipulation process is notified to the local processor 108, and this process is terminated.

  When the process proceeds from S902 to S903, a bit manipulation process is performed on the bit table on the RAM 104 having a large access delay. For this reason, the processing time is increased compared to the bit operation processing to the bit table in the internal memory 203 which advances the processing from S902 to S906. However, since the preload processing of the RBMC 112 is executed as in the sequence shown in FIG. 5 described above, the processing proceeds from S902 to S906, and the bit manipulation processing can be performed at high speed.

  According to the present embodiment, in the IP reassembling method described in RFC 791, the RBMC 112 can temporarily store the bit table in the internal memory 203 with a small access delay and perform the bit operation processing, so that the speed of the bit operation processing can be realized. .

  Also, since the execution of the IP packet reception process for the received fragment packet and the preload process of the RBMC 112 are performed in parallel, the time overhead for the RBMC 112 to store the bit table in the internal memory 203 can be ignored.

  Further, in the preload process, when there is no free space in the internal memory 203, control is performed to write back to the RAM 104 an IP reassembly bitmap table having a large arrival interval of fragment packets. Therefore, a bit table having a short fragment packet arrival interval, that is, a high frequency of IP reassembly processing can be preferentially stored in the internal memory 203, and bit table data transfer between the RAM 104 and the internal memory 203 can be performed. Can be reduced.

[Second Embodiment]
Next, a second embodiment according to the present invention will be described in detail with reference to the drawings. The configuration of the IP protocol device in the second embodiment is the same as that shown in FIGS. 1 and 2 described in the first embodiment. That is, the IP protocol processing apparatus 105 shown in FIG. 1 and the internal configuration shown in FIG. 2 are the same, and the reassembly bitmap controller (RBMC) 112 is also the same as the configuration of the first embodiment.

  In the first embodiment, the entire bit table data is stored in the internal memory 203 for some of the most recently processed IP reassembly among all the bit tables processed in parallel with the internal memory 203 of the RBMC 112. It was temporarily stored. However, in the second embodiment, the data temporarily stored in the internal memory 203 temporarily stores only partial data of all bit tables secured on the RAM 104.

  FIG. 10 is a diagram showing a state in which only partial data of all bit tables is stored in the internal memory 203. Reference numerals 1008 to 1014 denote that all bit tables of N IP reassembly are secured in the RAM 104. Further, it is shown that only partial data in the bit tables 1008 to 1014 are stored in 1001 to 1007 of the internal memory 203.

  In this example, partial data of the bit table 1008 with a reassembly ID of 0 is stored in 1001, and partial data of the bit table 1012 with a reassembly ID of M is stored in 1005.

  The storage method in the internal memory 203 is a direct map method. That is, in the internal memory 203, the storage location and size of a part of each bit table are defined as fixed, and the target data stored in each storage location is the range of 1K octets of each bit table secured on the RAM 104. Limited to

  In such a storage method of the internal memory 203, the memory size required for the internal memory 203 can be reduced by reducing the size of data to be stored. Further, the maximum data size required for one IP reassembly is stored as the data size stored in the internal memory 203. As a result, the data size transferred between the RAM 104 and the internal memory 203 in the preload processing of the RBMC 112 can be made smaller than when the entire bit table is transferred.

  Here, the size stored in the internal memory 203 is determined based on the MTU value of the network to which the IP protocol processing apparatus 105 is connected. For example, in the case of a general Ethernet (registered trademark), since the MTU is 1500, the maximum length of the payload carried by one IP packet is 1480 minus the minimum size of the IP header.

  Therefore, the fragment data carried by one fragment packet has a maximum length of 1480, and the necessary bit table data is 185 bits. That is, if bit data for 24 bytes (192 bits) is stored for each bit table, the data size is sufficient for IP reassembly of one fragment packet.

  In the first embodiment, the entire bit table is temporarily stored in the internal memory 203. Since the size required for one bit table is 1 Kbyte (8192 bits), the number of bits is approximately 40 times as many. It is possible to temporarily store the table.

  However, in the first embodiment, it is not necessary to transfer the bit table data from the RAM 104 when the bit table is already secured in the internal memory 203 by the preload processing of the RBMC 112. That is, when the same IP reassembly is processed continuously or when the execution frequency per unit time is large, there is an effect of reducing the number of bit table data transfers.

  On the contrary, in the method of the second embodiment, data transfer to the RAM 104 and the internal memory 203 occurs in each IP reassembly unless the fragment packet frequently has overlapping fragment data.

  However, the transfer size of the bit table data is small, and as described above, the IP reassembly is performed in parallel with the IP packet reception process of the local processor 108 and the preload process of the RBMC 112.

  Therefore, the time overhead required to transfer the bit table data between the RAM 104 and the internal memory 203 can be ignored, and the IP reassembly processing speed does not decrease.

  In addition, the number of IP reassembling processes that can be performed in parallel can reduce the memory cost of the internal memory 203 with a small access delay.

[Third Embodiment]
Next, a third embodiment according to the present invention will be described in detail with reference to the drawings. Note that the configuration of the IP protocol device in the third embodiment is the same as that of FIG. 1 described in the first embodiment. That is, the internal configuration of the IP protocol processing apparatus 105 shown in FIG. 1 is the same. However, the internal configuration of the reassembly bitmap controller (RBMC) described in the first embodiment is different.

  FIG. 11 is a diagram illustrating an example of the configuration of the RBMC in the third embodiment. Reference numeral 1100 denotes an RBMC in the third embodiment corresponding to the RBMC 112 shown in FIG. The RBMC 1100 includes a register 201, a resource management unit 202, a primary internal memory 1101, a bit processing unit 204, a DMA function unit 205, and a secondary internal memory 1102. In addition, the same code | symbol is attached | subjected to the thing similar to the structure shown in FIG.

  In the configuration of the RBMC 1100, the difference from the first embodiment is that the internal memory 203 includes two parts, a primary internal memory 1101 and a secondary internal memory 1102.

  FIG. 12 is a diagram illustrating a method of storing bit table data in the internal memory according to the third embodiment. In the RAM 104, all the IP reassembly bit tables 405 to 411 that are processed in parallel are secured. The secondary internal memory 1102 temporarily stores four bit tables among bit tables secured at arbitrary addresses in the RAM 104. In the example shown in FIG. 12, four bit tables 405, 408, 409, and 411 with reassembly IDs 0, 3, M, and N−1 are stored in locations 1209 to 1212 of the secondary internal memory 1102. . Further, partial data 1205 to 1208 of each bit table temporarily stored in the secondary internal memory 1102 are stored in 1201 to 1204 of the primary internal memory 1101.

  In this storage method, storage in the primary internal memory 1101 is a direct map method for the secondary internal memory 1102.

  In addition, the storage in the secondary internal memory 1102 is a fully associative method in bit table size units with respect to a bit table arranged at an arbitrary address on the RAM 104.

  In the configuration of the RBMC 1100 and the bit table data storage method, a bit table having a short fragment packet arrival interval and a large number of IP reassembly processes per unit time is preferentially stored in the secondary internal memory 1102. deep. By doing so, it is possible to reduce the transfer of bit table data to the RAM 104.

  Furthermore, since only a part of the data in the bit table is transferred between the primary internal memory 1101 and the secondary internal memory 1102, the overhead of data transfer can be minimized.

  Further, the primary internal memory 1101 and the secondary internal memory 1102 can transfer data at high speed within the RBMC 1100. Therefore, write-through control is performed so that data is written to the secondary internal memory 1102 at the same time that data is written to the primary internal memory 1101, and data between the primary memory and the secondary memory is recorded. Simplify synchronous control.

  Further, when the bit table to be processed by IP reassembly does not exist in the secondary internal memory 1102, only the data in the range to be processed is directly transferred from the RAM 104 to the primary internal memory 1101 so that the preload processing can be performed at high speed. Turn into. In this preloading process, it is not necessary to write back the data in the primary internal memory 1101 when performing the above write-through control.

  In this way, even when multiple IP reassembly is executed in parallel in the bitmap table type IP reassembly, the bit table is placed on a memory with a small access delay, and the bit operation processing is required to be accelerated and the memory is required. It becomes possible to suppress a large capacity. The present invention is suitable for a communication terminal device that performs IP protocol communication at high speed or in a TOE.

  According to the above-described embodiment, it is possible to increase the speed of IP reassembly by executing the bit manipulation processing of the bit table on the data held in the second storage unit with a small access delay.

  Also, all the IP reassembly bit tables that are in the process of processing are arranged in the first storage means, and only the bit table data for performing bit operations is placed in the second storage means with a small access delay, so that the second The storage capacity required for the storage means can be reduced. For example, an SDRAM is used as the first storage means and an SRAM is used as the second storage means. In general, an SRAM having a high memory cost is mounted in a small capacity, and an inexpensive SDRAM has a large capacity capable of storing at least all bit tables. By implementing it, memory cost can be reduced. That is, the number of IP reassembly processes that can be processed in parallel can be increased at a low memory cost.

  Further, bit table data for performing bit manipulation processing is prepared in the second storage means in advance of processing start, and IP reassembled bit table data having a short fragment packet arrival interval is preferentially stored in the second storage means. Hold. As a result, the time overhead of data transfer occurring between the first storage means and the second storage means can be reduced on average, so that the processing speed of IP reassembly can be increased.

  Even if the present invention is applied to a system constituted by a plurality of devices (for example, a host computer, an interface device, a reader, a printer, etc.), it is applied to an apparatus (for example, a copying machine, a facsimile machine, etc.) comprising a single device. It may be applied.

  In addition, a recording medium in which a program code of software for realizing the functions of the above-described embodiments is recorded is supplied to the system or apparatus, and the computer (CPU or MPU) of the system or apparatus stores the program code stored in the recording medium. Read and execute. It goes without saying that the object of the present invention can also be achieved by this.

  In this case, the program code itself read from the computer-readable recording medium realizes the functions of the above-described embodiments, and the recording medium storing the program code constitutes the present invention.

  As a recording medium for supplying the program code, for example, a flexible disk, a hard disk, an optical disk, a magneto-optical disk, a CD-ROM, a CD-R, a magnetic tape, a nonvolatile memory card, a ROM, or the like can be used.

  In addition, by executing the program code read by the computer, not only the functions of the above-described embodiments are realized, but also the following cases are included. That is, based on the instruction of the program code, an OS (operating system) running on the computer performs part or all of the actual processing, and the functions of the above-described embodiments are realized by the processing. .

  Further, the program code read from the recording medium is written in a memory provided in a function expansion board inserted into the computer or a function expansion unit connected to the computer. After that, based on the instruction of the program code, the CPU of the function expansion board or function expansion unit performs part or all of the actual processing, and the function of the above-described embodiment is realized by the processing. Needless to say.

It is a block diagram showing an example of a structure of the network communication apparatus in 1st Embodiment. It is a block diagram showing an example of composition of RBMC in a 1st embodiment. It is a figure for demonstrating the bit operation process of the bit table in IP reassembly. It is a figure which shows the state which manages the bit table of IP reassembly by an IP reassembly identifier. It is a figure which shows the time processing sequence of DMAC111, RBMC112, and the local processor 108 in IP reassembly. It is a figure which shows the format of the IP header of an IPv4 (IP version 4) packet. 6 is a flowchart showing processing of the local processor 108 (S505 to S508 in FIG. 5). It is a flowchart which shows the preload process of a bit table. It is a flowchart which shows the bit operation process of a bit table. It is a figure which shows the state which stores only the partial data of all the bit tables in the internal memory. It is a figure which shows an example of a structure of RBMC in 3rd Embodiment. It is a figure which shows the storage method to the internal memory of the bit table data in 3rd Embodiment.

Explanation of symbols

101 System bus 102 CPU
103 ROM
104 RAM
105 IP protocol processor 106 Local bus 107 Bus bridge 108 Local processor 109 Local processor 110 Local RAM
111 DMA controller (DMAC)
112 Reassembling bitmap controller (RBMC)
113 Communication timer 114 CAM
115 Cryptographic Processing Unit 116 Communication Control Unit 117 Network 201 Register 202 Resource Management Unit 203 Internal Memory 204 Bit Processing Unit 205 DMA Function Unit

Claims (7)

An IP protocol processing device for reassembling the IP packet based on a bit table for managing the reception status of the fragmented IP packet,
First storage means for storing a plurality of bit tables;
Second storage means for storing a part of the plurality of bit tables;
Transfer means for transferring the bit table to be reassembled from the first storage means to the second storage means when performing the reassembly based on the bit table;
Processing means for reassembling the fragmented IP packet based on the bit table transferred by the transfer means to the second storage means;
An IP protocol processing apparatus comprising:   The IP protocol processing apparatus according to claim 1, wherein the second storage unit has an access delay smaller than that of the first storage unit.   The apparatus further comprises means for controlling the transfer means to preferentially transfer the target bit table to the second storage means when the arrival interval of the fragmented IP packet is short. Item 3. The IP protocol processing device according to Item 1 or 2.   4. The IP protocol processing apparatus according to claim 1, wherein the transfer unit preferentially transfers data included in the bit table to the second storage unit. 5. Based on a bit table that has first storage means for storing a plurality of bit tables and second storage means for storing a part of the plurality of bit tables, and that manages the reception status of fragmented IP packets A processing method of an IP protocol processing device for reassembling the IP packet,
A transfer step of transferring the bit table to be reassembled from the first storage means to the second storage means when performing the reassembly based on the bit table;
A processing step of reassembling the fragmented IP packet based on the bit table transferred to the second storage means in the transfer step;
A processing method for an IP protocol processing apparatus, comprising:   The program for functioning a computer as an IP protocol processing apparatus of any one of Claims 1 thru | or 4.   A computer-readable recording medium on which the program according to claim 6 is recorded.

Images